Stock material or miscellaneous articles – All metal or with adjacent metals – Plural layers discontinuously bonded
Reexamination Certificate
1999-10-08
2001-05-29
Dixon, Merrick (Department: 1774)
Stock material or miscellaneous articles
All metal or with adjacent metals
Plural layers discontinuously bonded
C428S212000, C156S308400, C156S313000, C156S325000, C156S307300
Reexamination Certificate
active
06238805
ABSTRACT:
TECHNICAL FIELD
The present invention relates generally to joining dissimilar materials, and, more particularly, to a low-stress interface between materials having different thermal coefficients of expansion and method for fabricating same.
BACKGROUND OF THE INVENTION
Microsystems use the properties and movement of light, mechanical parts, gases, liquids, and electrons on a sub-millimeter scale to obtain useful functionality in optical, chemical, physical, biological, and electronics applications The joining of dissimilar materials in such systems poses challenges in fabrication, storage, and functionality due to bending, stress, and creep at the interface between such materials. Such an interface region is typically an active part of the device, not merely a passive mechanical joint.
As an example of a microsystem, commonly assigned U.S. Pat. No. 5,699,462 to Fouquet et al., which is hereby incorporated by reference, describes a microfluidic optical switch element located in a liquid-filled trench etched in glass at the intersection of two planar optical waveguide elements. The filling liquid has an index of refraction that matches that of the waveguide elements. If no gas-phase bubble is present at the intersection of the two waveguide elements, incoming light from either waveguide element continues through the trench on a straight-line path and exits on the opposite side of the trench into the continuation of that waveguide element on the opposite side of the trench. However, if a gas-phase bubble is present in the trench, the trench then has an index of refraction much lower than that of the waveguide elements, so that incoming light in one waveguide reflects at the surface of the bubble and is diverted into the other waveguide. The gas-phase bubble can be introduced and removed by electrical resistor heating means to perform optical switching.
The above-described optical switch element may be fabricated into an array of such switches, for example with 32 input channels, which can be switched to any of 32 output channels. This arrangement is referred to as a 32×32 switch, and contains 32×32=1,024 individual switch elements. The waveguides and associated trenches of such a switch are typically fabricated in a planar glass substrate with mechanical properties substantially identical to those of fused silica, most particularly with a very low coefficient of thermal coefficient of expansion (TCE) of 0.55 parts per million per degree Celsius (0.55 ppm/° C.). A ribbon cable of 32 single-mode optical fibers, typically on 250 micrometer (250 &mgr;m) centers can be attached to each of four edges of the glass substrate to provide input and output channels. Alignment of each single-mode fiber core to each planar waveguide element should occur with sub-micrometer accuracy to provide low insertion loss.
The resistive heating means of the 32×32 switch is typically a planar array of thin-film electrical resistors on a planar silicon integrated circuit substrate, most particularly with a TCE of 2.6 ppm/° C., which is low in comparison to most engineering materials but is substantially larger than that of fused silica.
To fabricate the complete 32×32 switch, the planar silicon substrate is attached to the planar glass substrate using some hermetic sealing means so that the resulting device can then be filled with a liquid whose index of refraction matches that of the planar waveguide elements. Subsequently, optical fiber ribbon cables can be attached to the edges of the planar glass substrate in an optically-active mode in which the optical transmission through the planar waveguide elements can be evaluated after the fiber ribbon cables are positioned, but before they are permanently attached. In order to achieve sub-micron accuracy of the alignment of each fiber core to each waveguide element, it is necessary to first obtain sub-micrometer bowing of the planar surface of the glass substrate along each edge of that substrate. Unfortunately, the mismatch between the TCE of the glass and that of the silicon tends to cause bowing of the composite device of at least several micrometers.
Such bowing, when it occurs in a device comprising two metal layers of different TCE, is referred to as a bimetal effect. By extension, such an effect occurring between any two materials of different TCE is also called a bimetal effect.
The bimetal effect between glass and silicon not only causes bowing, but also causes stress at the interface between the two materials as temperature varies. If the glass and silicon are joined to each other by solder at a soldering temperature of, for example, 165° C., where the hot solder freezes to form a solid, then, as the bimetal pair cools toward room temperature the silicon contracts more than the glass. The result is that the glass/silicon pair becomes concave on the silicon side and convex on the glass side, and the solder interface between the two materials experiences shear stress.
To illustrate this effect, shown in
FIG. 1A
is a cross-sectional schematic view illustrating a conventional microsystem
11
including glass portion
12
joined to silicon portion
14
via solder bond
16
. When heated to soldering temperature, both glass portion
12
and silicon portion
14
remain flat with respect to each other.
FIG. 1B
is a cross-sectional schematic view of the microsystem
11
of
FIG. 1A
after cooling to room temperature. As the microsystem
11
begins to cool from soldering temperature silicon portion
14
contracts more than glass portion
12
, wherein the microsystem
11
becomes concave on the silicon side and convex on the glass side. This bowing is a result of the above-mentioned bimetal effect.
FIG. 1C
is a cross-sectional schematic view illustrating the microsystem
11
of
FIGS. 1A and 1B
in which a smaller solder bond
16
is used to join glass portion
12
and silicon portion
14
. As the microsystem
11
cools, the gap
18
between glass portion
12
and silicon portion
14
varies in thickness along the planar interface between glass portion
12
and silicon portion
14
. The TCE difference between glass portion
12
and silicon portion
14
make the use of these two materials inadequate for a microsystem in to which the gap
18
should remain constant.
Typically, devices fabricated using this microsystem technology are required to survive temperatures as low as −40° C. during shipping and storage, and at such low temperatures the solder bond can be expected to fail. Even if outright failure of the bond does not occur, creep of the solder during long-term temperature excursions can lead to lateral and vertical displacement of the heating resistors relative to the waveguide crosspoints.
In addition, and more problematically, it is desirable to form the bond between the glass substrate and the silicon substrate only as a peripheral ring around the active 32×32 array of the two substrates. The use of a peripheral seal, in contrast to a large area seal, maximizes the volume in which microfluidic transport of gas and liquid can occur, and minimizes the structural complexity at each switch location within the 32×32 array. The necessity to include additional solder bond regions within the peripheral solder bond region would complicate the design and fabrication of such arrays, and would increase the cost of fabrication. However, because such a peripheral seal has only a small bonding area between glass and silicon, the stress (force per unit area) in the bond is greater than it would be for a large-area bond, and failure of the bond during temperature excursions becomes even more likely
Further, it is desirable to maintain a uniform spacing between the glass substrate and the silicon substrate over all of the 1,024 switch elements in the 32×32 array of switches. The fluid resistance per unit length between two adjacent flat plates varies as the third power of the spacing between the plates, so that small variations in spacing between two plates can produce large variations in fluid resistance. In the
Agilent Technologie,s Inc.
Dixon Merrick
LandOfFree
Low-stress interface between materials having different... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Low-stress interface between materials having different..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Low-stress interface between materials having different... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2563759